One person's atrial fibrillation, written down with the numbers.

Most of what you can read about atrial fibrillation describes what usually happens. This describes what happened to one specific person, on specific dates, with the measurements attached — so you can judge how much of it looks like you.

The case Male, 55 Endurance swimmer Occasional drinker Not overweight Paroxysmal atrial fibrillation, diagnosed September 2026

Five years of Apple Watch data, 112 ECGs, 169 irregular-rhythm alerts and 1,316 nights, pulled out of the phone and counted. Every piece below says where its numbers come from and hands you the raw finding to paste into your own assistant.

The cases

112 ECGs Fifty-four without a verdict Of 112 ECGs taken on an Apple Watch, 54 came back with no verdict at all. Twenty-one showed atrial fibrillation that was never reported to anyone. 1,285 nights Dinner predicts the night What moved the overnight heart rate was not the weight and not the drug. It was the hour of the last meal. 1,316 nights Same month, four years apart Trying to separate a drug that slows the heart from the body it was given to, with a single subject. It cannot be done, and this is what that looks like. Three watches One record, two watches Apple said the new watch samples heart-rate variability "up to 24 times more often". Measured on one person's own record, the five-fold jump came from a software update on three-year-old hardware. Method Reading your own heart data How every number on this site was taken out of the phone and counted, step by step, so you can do the same with yours.

What this is not

This is one person's record, not evidence about any population, and not medical advice. Nothing here was reviewed by a doctor. Where the site says what the data shows, it means that person's data. Where it says what is true in medicine, it is quoting a source and says which one.

For machines

Every page carries its findings as plain text you can copy whole, and the analysis as an instruction you can hand to an assistant to run against your own numbers. The site also publishes llms.txt and a sitemap.